Forthcoming in Mind & Language
In Defence of a Doxastic Account of Experience
Kathrin Glüer
Today, many philosophers think that perceptual experiences are conscious mental
states with representational content and phenomenal character. Subscribers to this
view often go on to construe experience more precisely as a propositional attitude
sui generis ascribing sensible properties to ordinary material objects. I argue that
experience is better construed as a kind of belief ascribing ‘phenomenal’
properties to such objects. A belief theory of this kind deals as well with the
traditional arguments against doxastic accounts as the sui generis view. Moreover,
in contrast to sui generis views, it can quite easily account for the rational or
reason providing role of experience.
--------------------------------------------------------------------------------
According to an account currently very popular in the philosophical literature,
perceptual experiences are conscious mental states with representational content and
Parts of this material have been presented at NYU, the CUNY Graduate Center, LOGOS in Barcelona,
the Perception and Introspection workshop in Glasgow, the Czech Academy of Sciences, and Lund,
Gothenburg, Umeå and Oslo Universities. I would like to thank participants of all these occasions for
helpful comments, especially Paul Boghossian, Michael Devitt, Kati Farkas, Manuel Garcia-Carpintero,
Olav Gjelsvik, Paul Horwich, Max Kölbel, Marie Lundstedt, Helge Malmgren, Genoveva Marti, Teresa
Marques, Fiona Mcpherson, Jim Pryor, Sven Rosenkranz, Camilla Serck-Hanssen, Susanna Siegel, Barry
Smith, Pär Sundström, and Dag Westerståhl. Special thanks, as always, to Peter Pagin and Åsa Wikforss.
Special thanks also to Tim Crane and an anonymous referee for this journal. Research funded by the
Swedish Research Council VR (project no. 2005-869).
Address for correspondence: Dept. of Philosophy, Stockholm University, 106 91 Stockholm, Sweden
Email: [email protected]
1
distinctive phenomenal character. If we accept such an account, we should be able to
provide answers to the following two questions: First, the ‘state question’: What kind of
a mental state is experience? And second, the ‘content question’: What form does its
content take?
Both of these questions, it seems to me, have been somewhat neglected in recent
debates. These debates have focused instead on the relation between phenomenal
character and representational content, and the kind of content experiences have. Thus,
we have seen the ‘qualia-wars’ and the search for a viable notion of non-conceptual
content. These clearly are intriguing issues, but they have diverted attention from the
more fundamental concerns mentioned above. The question I shall be most concerned
with in this paper is the state question: What kind of mental state is experience?
However, as we shall see, answering this question directly implicates the content
question, the question what form the content of experience takes.
Most people seem to agree on at least the following negative answer to the state
question: Experiences are not beliefs.1 And indeed, there are well-established, prima
facie convincing arguments against what we might call ‘belief theories’, or ‘doxastic
accounts’, of experience. The most well known of these arguments concern the so1
As far as I can tell, the position that experience is a kind of belief is presently not occupied by
anyone (but me). Armstrong, 1968, used to hold a version of the belief theory, as did Pitcher, 1971.
However, they never identified perception, or experience, as a kind of belief; rather, they conceived of
perception as an event, the event of forming a belief, or the inclination, or disposition to do so
(Armstrong, 1968, ch. 10). In the end, Armstrong settled for a counterfactual analysis of perception as
‘the acquiring of a potential belief: We come to be in a certain state which would be a belief-state but for
the inhibiting effects of other, contrary, beliefs’ (1968, 223). Clearly, such a state is not a belief. The last
to have defended a claim at least closely related to the claim that experience is a kind of belief is Craig
(1976) who held that ‘sensory experiences actually are judgements about our environment’ (1976, 8).
2
called ‘modularity’ or ‘belief-independence’ of experience. On the assumption that
experience nevertheless is a state representing the world as being a certain way, many
conclude that experience must be a propositional attitude sui generis.2
In this paper, I shall argue that these conclusions are at least hasty. As far as I can
see, the anti-doxastic arguments on the market all presuppose a certain construal of the
contents of experience. They all hinge on the premise that experiences take what I shall
call ‘naive contents’: They ascribe sensible properties like redness or roundness to
ordinary material objects. Though intuitive, such a ‘naive semantics’ is by no means
mandatory. I suggest replacing it by a ‘phenomenal semantics’, a semantics according to
which experiences do not ascribe sensible properties like redness or roundness to
ordinary material objects, but ‘phenomenal’ ones: properties like looking red or looking
round. Equipped with such a semantics, a doxastic account of experience is not only
back in business, but even outperforms any sui generis account, arguably its strongest
competitor.
I shall proceed as follows: In section 1, I shall outline both a basic characterization of
experience assumed as common ground in this paper, as well as the more precise
version of it that is standard in the literature. In section 2, I shall outline the most
important arguments against doxastic construals of experience. These arguments
concern the ways in which experience relates to empirical belief: Experience both
shows a characteristic independence from, while at the same time providing justification
or reasons for, empirical belief. These arguments point towards sui generity, but depend
on the assumption that experience has a naive semantics. In section 3, I shall suggest an
2
Sui generis accounts have been more or less explicitly suggested by, amongst others, Gareth Evans,
Tim Crane, John McDowell, Fred Dretske, Michael Tye, Jim Pryor, and Susanna Siegel.
3
alternative ‘phenomenal semantics’ for experience, and show how it brings the belief
theory back into business vis á vis ‘belief independence’. In section 4, I shall then
argue that, equipped with a phenomenal semantics, doxastic accounts have a decisive
advantage over sui generis accounts when it comes to accommodating the rational role
of experience.
1. The Standard Account of Experience
Here is a basic characterization of perceptual experience that is widely accepted today:
(E)
A perceptual experience
i)
is a conscious mental state,
ii) with a representational content, and
iii) a distinctive phenomenal character.
In this paper, I shall neither challenge, nor defend (E). Rather, I shall take (E) as
common ground that all parties to this particular discussion agree on.3 What I shall be
concerned with is the further analysis of i) and ii), more precisely, the answers
subscribers to (E) should give to the state and content questions as I have formulated
them above: the questions what kind of mental state experience is, and what form its
contents take. And I am going to challenge the most widely accepted answers to these
questions: These are the claim that the representational content of experience takes a
3
The following considerations thus do not engage accounts of experience according to which these do
not have representational content (as found, for instance, in Campbell 2002, Martin 2002, Travis 2004,
Brewer 2008); for very instructive recent discussion see Siegel 2005.
4
certain, ‘naive’ form, and the claim that experiences are not beliefs. Let’s start with the
content question.
Experiences, many people these days think, are about things out there, about the
world of ordinary material objects. And intuitively, they ‘tell’ us something about that
world. What they ‘say’ might be the truth, but it doesn’t have to be. ‘A perceptual
experience’, Christopher Peacocke, among many others, writes, ‘represents the world
as being a certain way’ (1992, 61). A perceptual experience, that is, ‘tells us’ that the
world fulfils a certain condition, for instance the condition of containing a red tomato or
a pink elephant right in front of us. The intuition that experiences ‘tell us’ such things
can thus be partly accounted for by construing them as having representational
contents: Contents with conditions of truth or correctness. In this very basic and
innocent sense, saying that experiences have representational content is saying that they
have propositional content.4
Just saying that experience has representational, or propositional, content leaves open
what I have called the ‘content question’ above, however: The question what form this
content takes. The arguably most important issues here are: What are the objects of
4
This is innocent in so far as it does not yet tell us as what kind of propositions the contents of
experiences are best construed, precisely what kind of relation their subjects are required to stand in
towards them, or what kind of medium of representation experience employs; these are questions I shall
remain neutral on. To the extent that the question whether experience content is conceptual or nonconceptual turns on any of these issues, I shall thus remain neutral on that question, too. There is a sense
of conceptual, however, on which beliefs, by definition, have conceptual content (Byrne (2005) calls this
the ‘pleonastic sense of concept’). In this sense, defending a belief theory of experience amounts to
construing experiences as having conceptual content. I do not think, however, that this, by itself, commits
the belief theorist to take a stand on any of the aforementioned issues. Thanks to Tim Crane for helping
me to get clearer on these issues.
5
experience? What kind of properties do experiences ascribe to these objects? And are
these contents singular or general? Since we are concerned with assigning truth
conditions to representations here, I shall call any answer to these issues a ‘semantics’
for experience.5
As already hinted, there is some consensus on the content question amongst
subscribers to (E). That is, there is widespread agreement that the objects of experience
are ordinary material objects such as chairs, books, and (most important from an
evolutionary perspective) fruit. There also is quite some agreement on the kinds of
properties that experiences ascribe to such objects: These are the so-called sensible
properties (or relations), properties like redness or roundness.6 I shall call any semantics
that accepts these two claims a ‘naive semantics’:
(NS)
Experiences have contents of the form x is F, where x ranges over ordinary
material objects, and F over sensible properties.7
5
Note that this does not commit me to a lingua mentis. The assumption is merely that as soon as there
are representational contents, these contens are represented somehow, in some medium of representation.
That alone is enough to justify talking about ‘semantics’ here; no assumptions about that medium’s
character or structure are required.
6
Consensus ends when it comes to distinguishing the sensible from the non-sensible properties. The
relevant question here is which properties an object can look to have, in the phenomenal sense of ‘look’
(cf. Jackson 1977: 33). For our present concerns, the precise answer to this question is of no further
importance.
7
Strictly speaking, and on the assumption that there are sensible relations, (NS) should be replaced by
(NS’)
Experiences have contents of the form F(x1, ..., xn), where xi (1≤i≤n) ranges over material objects
and F over sensible properties and relations.
6
Consensus ends when it comes to questions like whether (the most basic) experiential
contents are singular or general, object-dependent or object-independent. This, however,
will not matter here, and I shall remain (almost completely) neutral on these issues.8
What does matter here are the objects of experience, and the kinds of properties
ascribed to them. And while it certainly seems quite natural and intuitive to adopt a
naive semantics in response to the content question, it is important to note that (E) does
not commit us to such a semantics. As far as (E) is concerned, the objects of experience
might as well be sense data as ordinary material objects. Nor does (E) commit us to
sensible properties.
It might also be noted that (E) is neutral on questions concerning what we might call
the ‘foundational semantics’ of experience contents: By itself, (E) does not commit us
to take a stand on controversial questions regarding the determination of these contents.
For instance, you can subscribe to (E) regardless of whether you think that the content
of red-experiences is externalistically determined by the microstructural properties of
the red-looking things in the (local) environment. Both externalistic and internalistic
content determination are compatible with (E).
What about the ‘kind question’, then, the question what kind of a mental state
experience is? (E) tells us that experience is a conscious mental state. Moreover, it is
one of those that have representational or propositional content, so it clearly is some
kind of propositional attitude. But which? What kind of attitude is this? This question
needs to be distinguished from the question of whether (E), so to speak, captures all and
only the experiences. Even if it did that, even if it were extensionally correct, it would
8
That is why I shall mostly represent contents by means of open sentence schemas (like x is F); see
note 53 below, however.
7
not tell us which propositional attitude experience is. Take belief. (E) guarantees that
experience and belief is not just the same thing: Every experience has a certain
phenomenal character, and that is simply not true of every belief. Nevertheless, (E) is
perfectly compatible with experience being a kind, a subspecies, of belief. At the same
time, (E) is equally compatible with experience not being a kind of belief. And the same
would seem to hold for any other kind of attitude. Even if extensionally correct, (E) thus
leaves it completely open what kind of a propositional attitude experience is.
Nevertheless, there is widespread agreement at least on the following negative
answer to the kind question: Experience is not a kind of belief. But since there does not
seem to be any other kind of propositional attitude experience could plausibly be
subsumed under, many conclude that experience is a propositional attitude sui generis.
Incorporating naive semantics and sui generity into our basic account (E) of experience
results in the more precise version of (E) that is fairly standard today. As we shall soon
see, as long as we work with a naive semantics it is very plausible indeed to not only
hold that experiences are not beliefs, but to construe them as sui generis. In what
follows, I shall therefore treat sui generis accounts as the strongest standard competitors
to the doxastic account of experience I ultimately want to defend. First, however, I shall
take a closer look at the more or less traditional arguments against belief theories of
experience.
2. Against Belief Theories
2.1 Anti-Doxastic Arguments
The anti-doxastic arguments on the market can all be construed as reductio ad absurdum
arguments. Each of them focuses on a particular aspect or characteristic of the relation
8
between experience and empirical belief, arguing that construing experiences
themselves as a kind of belief cannot accommodate this aspect or characteristic: Trying
to combine the two results in absurdity.
Amongst these arguments, I shall roughly distinguish between two kinds. The first,
more traditional kind is phenomenological; it starts from certain phenomena readily
‘observable’ by any reflective subject of experience. As we shall see, the phenomena in
question all seem to support the idea that perception is, in a certain sense, modular or
‘belief-independent’, and I shall therefore call them the ‘modularity phenomena’. The
first kind of anti-doxastic argument concludes that doxastic accounts cannot
accommodate the modularity phenomena. The second kind of anti-doxastic argument
starts from a desideratum deriving not from the phenomenology, but from our pretheoretic notion of experience: According to this notion, experiences justify or provide
reasons for empirical beliefs. This second kind of anti-doxastic argument concludes that
doxastic accounts cannot account for experience’s reason-providing role.
Let’s start with the more traditional arguments from modularity phenomena. These
are often presented as arguments for the modularity of perception. What is meant here is
modularity in Fodor’s sense, where the perceptual system would be seen as a hardwired, informationally encapsulated module in the cognitive architecture of the mind.
What is interesting for our concerns, however, is only that part of modularity that is
concerned with, so to speak, putting experiences in a box of their own, a box different
from the belief box. For our purposes, that is, modularity is equally well captured by
what Evans called the the ‘belief-independence’ of experience (Evans, 1982, 123). We
shall have to take some care, however, not to describe the relevant phenomena in ways
that beg the question against doxastic accounts.
9
The probably most influential such argument might be called ‘the argument from
known illusion’. There are certain illusions that are extremely robust, the Müller-Lyer
illusion, for instance:
We all know that the two horizontal lines are of equal length. But knowing this does not
change the experience: One of the lines looks longer than the other, no matter what we
know or believe about their length. No amount of reflection can do anything about that.
It is thus very natural and plausible to think that a visual experience of a Müller-Lyer
diagram is independent of background belief in the following sense: Its content does not
vary or change with relevant background belief. Whether we know that we are subject
to an illusion or not, our experience has precisely the same content.
And now, we can run an argument to the conclusion that experiences are not beliefs.
Assume, for reductio, that experiences are beliefs. And assume that (at least some)
experiences are independent from background belief. Now, take someone, Peter, who is
subject to a known illusion: Peter is looking at a Müller-Lyer diagram and Peter has a
firm background belief that those two lines are of equal length. But they look as if one
of them was longer than the other. They look, that is, as if they were not of equal length.
By assumption, that is another belief. Thus, Peter ends up believing both that the two
lines are of equal length and that the two lines are not of equal length – a straight
contradiction. On the assumption that experiences are beliefs, independence from
background belief thus leads to the conclusion that subjects of known illusions have
contradictory beliefs.
10
This conclusion is bad, but maybe not yet clearly absurd. However, in cases of
known illusion the subject can be fully aware of both his experience and his background
belief. Peter is fully aware of believing what he, at the same time, knows to be false. No
amount of reflection can change this. For those still inclined to think that this is bad, but
not clearly absurd, here is what clinches the matter: Even perfectly rational subjects can
be subjects of known illusions. Peter is such a one: There is no plausible psychological
explanation whatsoever for his holding beliefs with very simple, clearly contradictory
contents, beliefs that he is fully aware of. This is absurd. Beliefs just don’t behave like
that.9 The argument from known illusion concludes that experiences are not beliefs.10
A rather similar argument might be called ‘the argument from falsely believed
illusion’. Consider the following example: You are in a store trying to choose a necktie.
The tie you are looking at looks green to you. However, you believe that the lighting in
the store is nonstandard and therefore withhold judgment as to the tie’s color. Assume
further that that background belief is false, and the lighting in fact is standard. On the
assumption that experiences are beliefs, you now are in the following situation: The tie
looks green to you. This is a belief, more precisely the belief that the tie is green. At the
same time, the background belief is supposed to have the effect of your withholding the
9
Note, that it therefore is not an option to construe the known illusion subject as compartmentalized.
Compartmentalization is a form of irrationality where contradictory beliefs are somehow ‘walled off’
from each other. Being subject to a known illusion is not sufficient for irrationality. In particular, there is
nothing ‘irrational’ about the lines looking of different length; given the structure of the diagram and
given that the subject’s visual system is in working order, this is what the lines ‘should’ look like.
10
This argument can be found already in Broad, 1923, and Armstrong, 1968. For more recent uses of
it, see for instance Jackson, 1977, 38ff and Crane, 1992, 149ff . See also Crane 2001, 150f where the
point is put in terms of Moore’s paradox: While asserting a content of the form I believe that p but not-p
is Moore-paradoxical, there is no such oddity in asserting ‘I experience that p but not-p’.
11
belief that the tie is green. Consequently, you end up simultaneously both believing and
not believing that the necktie is green.11 This is absurd. The argument from falsely
believed illusion concludes that experiences are not beliefs.
Another modularity phenomenon is provided by Tim Crane (cf. Crane, 1988; 1992,
150). He describes a type of illusion called the ‘waterfall illusion’. In this illusion, it
actually looks as if something was both moving and not moving at the same time. This
shows, Crane argues, that experiences can have outrightly contradictory contents: We
can, for instance, have an experience with the content x is moving and not moving. On
the assumption that experiences are beliefs it would therefore be possible to have beliefs
with outrightly contradictory contents. But this is absurd; no such beliefs are possible.
Therefore, the ‘waterfall-argument’ concludes, experiences are not beliefs.
The anti-doxastic arguments presented so far all use phenomenological data about
experiences either familiar to, or easily reproducable by, every reflective subject of
experience to derive absurd consequences from the assumption that experience is a kind
11
Strictly speaking, it is inessential here whether the background belief about the lighting is false or
not. All that matters for our purposes is that the belief that the tie is green is withheld. However, an
argument like this can be extracted from the use John McDowell (2004) makes of the necktie example
(originally from Sellars 1963, 143). There, McDowell argues that experience does not provide inferential
reasons for beliefs. Rather, having a (veridical) experience is having a certain sort of entitlement to belief,
an entitlement that cannot be overridden by background belief. In the necktie case, McDowell submits, no
belief that the necktie is green is formed, but the subject nevertheless is entitled to such a belief. And this,
of course, is only the case if the background belief about the lighting is false. But McDowell not only
argues that there is entitlement to the belief that the necktie is green in this case, he also concludes that the
experience, as ‘the entitling circumstance’, cannot be that belief: ‘that the entitling circumstance itself
needs to be distinguished from that belief is brought out by the fact that one can have the entitlement
without realizing that one does, and so without having the belief at all’ (2004, 215).
12
of belief.12 The last anti-doxastic argument I shall look at can be set up as a reductio ad
absurdum as well, but the premises it employs in combination with the assumption that
experiences are beliefs are of a different nature. The argument I have in mind might be
called ‘the stuttering inference argument’, and we find it in McDowell’s writings. The
stuttering inference argument turns on the rational or reason-providing role of
experience, and therefore requires some stage setting.
Like few others, McDowell has long insisted on the importance of experience’s
reason-providing or rational role. That experience provides us with reasons for
empirical beliefs, he argues in Mind and World (1994), is a condition on the very
possibility of having any mental states with empirical content. What he said in Mind
and World was interpreted by many as an attempt to extend a certain intuitive
conception of having reasons to experience. According to this conception, having
reasons for a belief that p is understood on an inferential model: Those reasons are such
12
Another phenomenological datum that is sometimes presented as evidence against doxastic
accounts is the characteristic ‘richness’ of experiential content. As Crane puts it: ‘The content of
experience is replete in a way the content of belief is not’ (2001, 151). But this conclusion clearly does
not follow from the mere datum of richness – unless we already assume that experiences are not beliefs.
The characteristic richness of experiential content might, or might not, be a good argument for the nonconceptual character of this content (as Evans was first to argue (1982, 229); but see for instance
McDowell (1994, 56ff) for arguments to the contrary). In that case, the belief theorist would have to say
that there are beliefs with non-conceptual contents (as, indeed, many proponents of non-conceptual
content (such as Tye (2002)) hold anyway). Another, arguably more plausible option, would be to
consider richness a matter of the medium of representation; Dretske, for instance, tries to account for
richness in terms of ‘analog’ (as opposed to ‘digital’) coding (1981, 135ff). But again, this would only be
an argument against a belief theory if it was already established that beliefs cannot employ analog coding.
13
that they could be used as premises in a valid (logical or material) inference to the
conclusion that p.13
In the ensuing discussion, however, McDowell pointed out that absurd consequences
arise from such an account of the rational role of experience. More precisely, absurd
consequences arise if we combine an ‘inferential’ conception of reasons with the
following observation about the strength of experiential reasons. Intuitively, an
experience as of p provides its subject with a rather strong reason for believing that p.
However, such a reason can be overridden by background belief. This is precisely what
happens when you believe, for instance, that lighting conditions are non-standard. In
general, experiential reasons can be overridden by background belief to the effect that
conditions are such that the senses are not to be trusted. Experiences, in other words,
seem to provide defeasible reasons for empirical belief.
Let’s call those empirical beliefs most directly supported by, or formed most directly
on the basis of, experience ‘basic perceptual beliefs’. Now, in order for experience to
provide defeasible ‘inferential’ reasons for such beliefs, the degree to which an
experience inferentially supports a basic perceptual belief must be less than 1. But, as
McDowell repeatedly points out, when it comes to experience and basic perceptual
belief, what we actually seem to be concerned with are two propositional attitudes with
the same content. The degree to which p evidentially supports p is 1. An inference from
13
Among those that understood McDowell in this way are Brewer, 1999, Wright, 1998, and Glüer
2004. Such an interpretation surely is at least invited when McDowell exemplifies the relevant relations:
‘rational relations, such as implication or probabilification’ (1994, 53).
14
p to p thus is not defeasible; there is no kind of background belief that could (rationally)
override such an inference.14 McDowell writes:
[W]hat matters is the rationality exemplified in judging whether things are thus and so in
the light of whether things are (observably) thus and so. The content of the item in the
light of which a judgement of this kind has its rational standing is the same as the content
of the judgment itself. The only inferences corresponding to the rational connection in
question would be of the “stuttering” form, “P, so P.” No doubt that inference-form (if we
allow it the title) cannot lead one astray, but its freedom from risk seems a quite unhelpful
model for the rationality of observational judgment (1998, 405; cf. also 1997, 161).
McDowell is quite right in complaining that, intuitively, basing beliefs on experience is
not ‘riskfree’ in the sense of experience, just by itself, providing indefeasible, or
conclusive, reasons: Surely, reasoning on the basis of experience does not stutter.
On the face of it, this is an argument against modelling the rational role of experience
by the inferential conception of reasons. However, very little needs to be added to make
the stuttering inference argument into an argument against a belief theory of experience.
For if experience itself is a kind of belief, then its rational role is precisely that of belief
providing reasons for (further) empirical beliefs. And I take it to be uncontroversial that
the model of choice here is (some version of) the inferential model.15 Consequently, the
14
Note that the point holds even if we think of experiential contents as typically (or even necessarily)
much richer than the contents of any perceptual beliefs. Experiential contents then could be understood on
the model of very long conjunctions, and basic perceptual beliefs as (necessarily, or typically) having
only some of the conjuncts as their contents. Nothing can defeat an inference from p & q to p.
15
There is controversy about the degree to which such reasoning should be construed externalistically.
Williamson (2000), for instance, argues that only those beliefs that are in fact knowledge provide
15
assumptions that experiences are beliefs, and that they provide reasons for further
empirical beliefs, lead to the absurd consequence that the relevant inferences would be
of the ‘stuttering’ kind. The stuttering inference argument concludes that experiences
are not beliefs.
All in all, it now looks as if doxastic accounts led to an impressive range of rather
absurd consequences: They cannot account for the modularity phenomena. Coming with
the inferential model of reasons ‘built in’, they might have some initial promise when it
comes to accounting for experience’s reason-providing role, but a closer look reveals
the absurd consequences of thus extending this model to experiences. These arguments
are very powerful. To my mind at least, both accommodating experience’s
independence from background belief and accommodating its reason-providing role
simply are desiderata on any satisfactory account of experience. So, if the anti-doxastic
arguments presented are good ones, we have every reason to think that experience is not
a kind of belief. Moreover, as we shall see next, there also would be very good reason to
conclude that experience is a propositional attitude sui generis.
2.2 On Representing the World as Being a Certain Way
Experience not only has representational content, it represents the world as being a
certain way. An experience with the content that p represents the world as being such
that p is true. Experience therefore is what I shall call a ‘strongly representational state’:
(SR)
A mental state is strongly representational iff
i)
it has a propositional content p, and
evidence. But this is a controversy about which states provide reasons or evidence, not about the
inferential nature of the reasons or the evidence provided.
16
ii)
it represents the world as being such that p is true.
I shall call a mental state ‘weakly representational’ iff i) is satisfied.
The distinction between weakly and strongly representational states is most easily
illustrated by means of examples such as desires, imaginings, assumings and
entertainings. Desires have representational content, but they do not represent the world
as being the desired way. Entertaining a proposition involves tokening a representation,
but does not amount to representing the world as being the entertained way. Making an
assumption, for instance to see what follows from it, or to prove that it has a certain
(logical) consequence, is not representing the world as being the assumed way.
Imagining the world to be a certain way is not representing it as being the imagined
way. But believing the world to be a certain way is representing it as being that way;
belief is the paradigmatic case of a strongly representational state.
Saying that experience represents the world as being a certain way thus is saying that
experiences are like beliefs in this respect: both are strongly representational states.
Experience is more like belief than like desire, or imagining, or entertaining a
proposition when it comes to representing the world as being a certain way. This much,
I take it, is uncontroversial.16 It is also overwhelmingly intuitive; it is the most natural
16
The distinction between weakly and strongly representational states more or less precisely coincides
with that between ‘semantic’ and ‘stative’ representation (Martin 2002, 386f), and that between
‘assertively’ and ‘non-assertively’ representing that p (Pryor 2005, 187f). Heck is clear about that this is a
matter of the attitude towards p, and distinguishes accordingly between ‘assertive, or presentational’ and
non-assertive attitudes (2000, 509). Strongly representational states could also be characterized in the
somewhat metaphorical terms of Searlian ‘direction of fit’: Strongly representational states have mind-toworld direction of fit, while desires and intentions have world-to-mind direction of fit (cf. Searle 1983, 7f;
imaginings and entertainings, I take it, have ‘the null direction of fit’ (Searle 1983, 8)). Not less
17
way for anyone subscribing to our basic account of experience (E) to accommodate that
aspect of the phenomenology of experience that has been called its ‘immediacy’.17
Visual experience, for instance, is immediate in the sense that ‘[v]isual phenomenology
makes it for a subject as if a scene is simply presented’ (Sturgeon 2000, 9). As Martin
points out, even if the subject suspects that he is having an illusion, or even that he is
hallucinating, ‘the fact that he draws that conclusion need not be sufficient to alter how
the situation strikes him phenomenologically: it may still seem to him as if things are
so’ (Martin 2002, 401; emph. mine). For anyone who thinks that experiences do have
propositional contents, the natural way of cashing out their immediacy is to say that
they represent or, as Searle would have it, ‘present’ the world as being such that that
content is true.
But once we thus classify experience with belief as strongly representational, the step
to thinking them a kind of propositional attitude sui generis is rather short, indeed. For
one thing, it is by no means implausible to think that beliefs and experiences are the
only strongly representational attitudes. And for another, even if they were not, there
simply does not seem any other attitude with quite the same, rather peculiar relation to
empirical belief. The underlying general idea here is to type attitudes by their functional
metaphorically, strongly representational states could be said to ‘aim at truth’. Thus Crane, for instance,
writes: ‘[P]erceptions have this in common with judgement, or the formation of belief: they “aim” at
truth’ (2001, 150).
17
Cf., among others, Martin 2002, 399; Searle 1983, 45; Sturgeon 2000, 9. Martin in fact suggests that
conceiving of experiences as what he calls ‘stative’ representation is the only plausible way of accounting
for immediacy available to the intentionalist (2002, 387f), where an intentionalist about perceptual
experience holds that experience can have the same propositional content across veridical, illusory and
hallucinatory cases. As far as I can see, the point, if good, would generalize to all accounts on which
experience is a propositional attitude.
18
roles. Compare beliefs and experiences: Beliefs and experiences are quite alike when it
comes to the ‘output’ side of their functional roles. Both ‘stand ready’ to make an
impact on further cognitive states.18 Nevertheless, experiences are distinct; what they do
not share with beliefs is the ‘input’ side of their functional role. Beliefs take input from
other strongly representational attitudes, i.e. further beliefs or experiences, but
experiences do not. They are caused by sub-personal cognitive processes.19
Thus far, the case for sui generity is rather weak, however. There is no obvious
reason to think that an attitude is sui generis just because it is, so to speak, the first of a
certain kind in the order of processing. That, in the order of processing, experiences are
the first personal and strongly representational states, does not just by itself provide us
with any reason to think that they are not beliefs. There is no reason in principle why it
would be impossible for personal level processing to start with beliefs.20
The case for sui generity only gets off the ground once we start describing the
relevant functional roles in terms of content. At this level, beliefs clearly and
distinctively are characterized by certain constraints of coherence: Individual beliefs do
not tolerate outrightly contradictory contents, and belief systems are subject to
constraints on their internal coherence. This induces certain behaviors in the face of
newly acquired information: Beliefs can be ‘overridden’, but for a belief to be
overridden means extinction. None of this seems to hold for experiences. This is
18
This ‘output’ side of their functional role, you might want to insist, is the naturalistic analogue, or
counterpart, of the reason-providing role we have been talking about above. What is distinctive of reasonproviding, of course, is that it is a content-sensitive relation.
19
Cf. for instance Tye 2002, 62f. That experience has this distinctive functional role is what he means
when he calls the content of experience ‘poised’.
20
As for instance Davidson held it to do, cf. Davidson 1983.
19
precisely what the modularity phenomena seem to show. If you know that the lines of a
Müller-Lyer diagram are of equal length, your experience ‘telling you’ otherwise is
overridden, but it does not cease to exist. Consequently, it looks as if the constraints on
internal coherence governing belief did not reach experience. Individual experiences
even seem to tolerate directly contradictory contents (witness the waterfall illusion).
Once we characterize functional role at the level of content, it therefore seems very
plausible to think that, as a matter of fact, the functional role of those states that are first
in the order of personal processing is distinctive, indeed. Experience, it thus seems, is a
strongly representational propositional attitude sui generis. The kind question cannot be
answered by subsuming experience under any other kind of propositional attitude; if we
want to know more about which kind of attitude experience is, what we should do
instead is carefully describe its functional role in counterdistinction to that of its ‘closest
cousin’: belief.
This whole line of anti-doxastic argument is very forceful. Nevertheless, I shall
spend the remainder of this paper arguing that there not only is a way for the belief
theory to make its comeback, but a way resulting in an account of experience superior to
sui generis accounts. For starters, I shall bring out an important hidden premise which
the whole anti-doxastic construction hinges on.
2.3 The hidden premise
To see that there is a way in which the belief theorist can deal with both the modularity
phenomena and ultimately even give a plausible account of the rational role of
experience, we first need to realize that there is a hidden premise in the anti-doxastic
arguments presented. In fact, there is the same hidden premise in all of them. This is the
assumption that experience has a naive semantics.
20
According to naive semantics, experiences take ordinary material objects and ascribe
sensible properties to them. And that is precisely what the beliefs relevant to the antidoxastic arguments do, too: They ascribe the same properties to the same objects. For
all of these beliefs it thus holds that they have the same contents as the relevant
experiences.21 In the argument from known illusion, the relevant belief is a background
belief that not p while the experience has the content p. Thus, contradiction. In the
stuttering inference argument, the relevant belief is a basic perceptual belief that p, and
the experience also has the content p. Thus, stuttering. Naive semantics is assumed in
the argument from falsely believed illusion, as well. It is only on this assumption that
the subject here ends up both having and not having the very same belief. And naive
semantics is assumed in the waterfall argument: If experiences ascribe sensible
properties to their objects, ‘waterfall contents’ are contradictory. To turn matters
around, it is only on the assumption of naive semantics that these anti-doxastic
arguments go through.
Moreover, naive semantics is crucial in the case for sui generity, too. As we saw in
the previous section, it is only once the functional role of experience is described at the
level of content that this case picks up speed. It is only on the premise that experiences
have the naive contents that enable them to contradict background beliefs (and even to
have contradictory contents) that their functional role appears so peculiar.
21
Note, again, that even if you think that experiential contents typically, or even necessarily, are much
richer than those of the relevant beliefs, it will still hold, on a naive semantics, that the contents of these
beliefs are parts of the experiential contents in question (see above, note 14). Such partial identity is
enough to generate the modularity phenomena, and thus sufficient to implicate naive semantics in their
generation.
21
The idea therefore is to give up this assumption. If we can find a plausible alternative
semantics for experiences, a semantics that neither induces identity of content between
suitably related experiences and (basic perceptual or background) beliefs, nor gives us
contradictory contents in waterfall situations, we might be able to resuscitate the
doxastic account of experience.
I do not mean to deny that the claim that experience has a naive semantics prima
facie is rather plausible, however. It definitely is what many philosophers – at least
those with a lot of training in contemporary philosophy of mind – would come up with
off the top of their heads. And it definitely is what many such philosophers find
intuitive. Not only that, it might also seem that intuition here is backed up by
phenomenology again. Just as introspection reveals its independence from background
belief, introspection is taken by many to show that experience is ‘transparent’: In having
an experience, its subject is not (directly) aware of any inner objects or qualities of the
experience itself. The only objects and properties she is (directly) aware of are external
ones.22 On the assumption that what you are thus aware of is what is represented by
your experience, you might then take transparency to back up naive semantics.23 In sum,
you might think that having a naive semantics is as non-negotiable a desideratum on an
account of experience as accommodating the modularity phenomena. I do not agree;
surely, assumptions about the precise form of the propositional content of a particular
kind of propositional attitude are fairly theoretical claims.24 The observation that certain
22
Cf. Harman 1990; Tye 2002, 47.
23
Thanks to an anomynous referee for helping me get clearer about the potential role of transparency
in this context.
24
I shall come back to the issue of transparency, however, once I have actually sketched an alternative
semantics.
22
experiences show a characteristic independence from background belief, on the other
hand, is something every reflective human subject can introspectively ascertain once in
a relevant situation. It is fairly clear, I think, that the intuitions behind naive semantics
are of lesser weight than the modularity phenomena. Consequently, it would be quite
wrong to reject the idea of an alternative semantics for experiences out of hand. Rather,
we should assess its overall theoretical merits once a positive suggestion for such a
semantics is on the table. Once we actually look at it, it might even turn out not to be so
very counterintuitive. Let me therefore in the next section sketch an alternative
semantics for experience.
3. Phenomenal Semantics and the Return of the Belief Theory
3.1 Phenomenal Semantics
We do not have to stray far in order to find a promising alternative to naive semantics.
We can hold on to the extremely intuitive claim that the objects of experience are
ordinary material objects. All we need to fiddle with is the properties (and relations)
experience ascribes to them. Take visual experience. What remains constant regardless
of background belief in the known or falsely believed illusion cases is the way things
look. The Müller-Lyer lines look to be of different length, even though we know
perfectly well that they are not. And in waterfall cases, an object looks to be moving
while at the same time it looks not to be moving. How things look is also what we cite
when we cite visual experiences as reasons; we say that we believed the tomato was red
because it looked that way, it looked red.
In ordinary speech, ‘looks’ is used in a number of different senses. The use we are
interested in here is the sense Jackson calls ‘phenomenal’ (1977, 33): This use of
23
‘looks’ is tied to terms for sensible properties (examples would be: ‘It looks red’ or ‘The
lower line looks longer than the upper line’), but that a material object x looks F to a
subject S neither implies that x is F, nor that S believes x to be F.25 Nevertheless,
looking F is a perfectly respectable property of x; it is what I would like to call a
‘phenomenal property’.26
In ordinary speech, we use the phenomenal looks-locution both to indicate a
propositional attitude, that of experience, and to indicate the content of that attitude. If I
report my experience by saying ‘The tomato looks red’, this report can be ‘translated’
into more technical philosophical lingo as ‘I am having an experience as of the tomato’s
being red’. On an account of experience working with a naive semantics, this is further
analyzed as my having an experience with the representational content the tomato is
red. What I would like to suggest instead is that this experience has the content the
tomato looks red. Experience, that is, does not ascribe sensible properties to ordinary
material objects, but phenomenal ones. It does not have a naive semantics, it has a
phenomenal semantics. In general, visual experience has contents of the form x looks F.
And analogously for the other senses.27
25
This is the use of ‘appear words’ Chisholm calls the ‘noncomparative use’ (1957, ch. 4, esp. 50ff).
26
Phenomenal properties in this sense might for instance be dispositions to cause certain reactions in
minded creatures (the dispositional properties Shoemaker calls ‘phenomenal properties’ (1994) or
‘appearance properties’ (2000) would be examples here), or they might be the categorical bases of such
dispositions. They might also be properties that consist in an object’s actually causing the reactions in
question.
27
In general, we might say, a phenomenal semantics construes experience as having (phenomenal)
appears- or seems-contents. I do not think, however, that we should construe all experiences,
independently of their sense modality, as having contents of the form x appears F, but rather as having
modality-specific contents, contents of the form x looks F, x sounds G, etc. Experience thus carries some
24
Now, I, for one, cannot find this suggestion violently counterintuitive. Nor do I think
it really has much quarrel with transparency. Transparency has it that the only objects
and properties we are (directly) aware of in experience are external objects and their
properties. On the assumption that what we are thus aware of is what is represented in
experience, transparency supports the claim that what is represented in experience is
nothing but external objects and their properties. Both a naive and a phenomenal
semantics respect this dictum. Consequently, transparency would favor a naive
semantics over a phenomenal one only if introspection also told us that the properties
represented in experience are sensible properties like redness or squareness, rather than
phenomenal ones.
But just how plausible is it to expect introspection to deliver this result? In what way
would the phenomenal character of an experience differ if it had the content x looks red
instead of x is red? That differences of this kind are detectable by introspection is
questionable already within one and the same ‘semantic framework’,28 but in order to
cut any ice against phenomenal semantics, it would have to hold even across such
frameworks. In order to defend phenomenal semantics, we have to argue that
information about its own sense modality. I think that this is phenomenologically correct; there is a
phenomenal difference between an object’s looking square and its feeling square, for instance. On the
suggested phenomenal account of the representational content of experience there is, therefore, no
problem about common sensibles even if representationalism is true, that is, even if representational
content determines phenomenal character. With respect to the problems that are the topic of this paper,
however, I don’t think it matters whether the phenomenal semantics works with modality-specific or
modality–inspecific contents; both versions of phenomenal semantics deal equally well with these
problems.
28
Disjunctivists, for instance, typically think that a verdical experience as of p and a corresponding
hallucination can be introspectively indistinguishable, but nevertheless do not have the same content.
25
phenomenal contents have precisely the same phenomenology as naive contents: They
are phenomenally equivalent. This is very plausible. Take an experience as of something
red. Such an experience has a distinctive phenomenal character or quality Q. It is
intuitively very plausible that an experience with the content that x is red has Q. But it
is equally plausible that an experience with the content that x looks red has precisely
this phenomenal character. Having Q therefore is equally compatible with either
content; it does nothing to decide between them.29
In what follows I shall, however, simply grant that adopting a phenomenal semantics
for experience remains somewhat counterintuitive. For even if that is so, I claim, once a
phenomeal semantics is combined with a doxastic account of experience, the overall
advantages and unifying power of the resulting position amply make up for it. This
claim will of course have to remain somewhat programmatic in this paper. And so will
the suggested semantics itself. I shall not develop it any further here; in this paper, I
shall only show how working with phenomenal properties in the semantics of
experiences allows the belief theorist to deal with the anti-doxastic arguments. In
particular, I shall argue in the remainder of this section that combined with a
phenomenal semantics, a doxastic account of experience provides as good an account of
the modularity phenomena as a sui generis account.
29
Note, that the phenomenal semanticist remains free to endorse (weak) ‘representationalism’, i.e. the
claim that phenomenal character supervenes on, or is determined by, the representational content of
experience. Note, too, that within a given semantic framework it might also hold that phenomenal
character determines representational content.
26
3.2 The Return of the Belief Theory
The modularity phenomena arise in sitations where naive contents would result in
absurd beliefs, if experiences were beliefs. Faced with these phenomena, the sui generis
theorist recommends to construe experiences as a kind of state that is distinct from
belief precisely in being tolerant of naive contents in these situations. In short, the sui
generis theorist responds to the modularity phenomena by fiddling with the kind of state
experience is. Looking at the situation in the abstract, it should not be very surprising
that there is another, complementary move available here that consists in fiddling with
the contents of experience instead. Adopting a phenomenal semantics for experience is
precisely such a move.
Let’s quickly go through the relevant cases to see how adopting such a semantics
deals with the modularity phenomena. First, known illusion: Peter, our known illusion
subject, knows that the lines of the Müller-Lyer diagram he is looking at are of the same
length, but they nevertheless look to be of different length to him. On a phenomenal
semantics, this amounts to an experience with the content that the lines look to be of
different length, and a background belief that they are of the same length. There is no
contradiction between the contents of the two states (cf. Hamlyn, 1994, 253), and, thus,
no pressure towards the conclusion that the experience is not a belief. Next, falsely
believed illusion: John falsely believes that the lighting is non-standard. He refrains
from forming the basic perceptual belief that the necktie he is looking at is green, but
the necktie nevertheless looks green to him. On a phenomenal semantics, this amounts
to his having an experience with the content that the necktie looks green, but not having
having the belief that it is green. No support for the conclusion that the experience is not
a belief is to be had from this. Last, but not least, the waterfall: Tim, our waterfall
subject, has a visual experience as of an object x’s both moving and not moving at the
27
same time. On a phenomenal semantics this can be construed as having an experience
with a content of the form x looks F and x looks not-F. The phenomenal content is not
contradictory, and there is thus no pressure towards the conclusion that a ‘waterfall
experience’ is not a belief. 30
In sum, adopting a phenomenal semantics for experiences removes the pressure that
the modularity phenomena, on the assumption of a naive semantics, exert towards the
conclusion that experience is not a kind of belief. If a doxastic account of experience is
combined with a phenomenal semantics, it has no difficulties accommodating
experience’s independence from background belief. Not only is there no contradiction
between both believing that the Müller-Lyer lines look to be of different length and
believing that they, in fact, are of the same length; on the assumption that experience is
a kind of belief, having both of these beliefs nicely captures what intuitively is going on
in a case of known illusion. A belief about how things look is precisely what is
independent of background belief to the effect that things in fact are different from what
they look to be in the situation in question. On a phenomenal belief theory, the so-called
‘belief independence’ of experience, or more precisely, the modularity phenomena
underlying this claim, can thus be very nicely and elegantly captured as experience’s
independence from background belief. As far as the modularity phenomena are
concerned, adopting a phenomenal semantics puts the belief theory back into business.
30
What would be contradictory, of course, is a content of the form x looks F and x does not look F.
But the phenomal semanticist clearly does not have to construe a ‘waterfall experience’ as having a such a
content. Moreover, doing so would not capture the phenomenology of waterfall illusions at all; the
waterfall object looks to be moving and it looks to be not moving. In general, not looking F is not a
phenomenal property, while looking not-F can be (where F is a suitable sensible property).
28
Now, of course the sui generis theorist could adopt a phenomenal semantics, too.31
As far as options in logical space are concerned, the kind question might well be
orthogonal to the content question. But it is far from clear what motivation the sui
generis theorist could have for adopting a phenomenal semantics. As far as such a
motivation is supposed to come from the modularity phenomena, it is certainly not
required to perform both of the defusing moves that are available in reaction to it.
Rather, their motivating power would seem to be exhausted by making one or the other.
Thus, adopting a sui generis account in combination with a naive semantics, or a
doxastic account in combination with a phenomenal semantics seem to be equally
effective, and in a certain sense complementary, responses to the phenomena in these
cases. As far as the modularity phenomena are concerned, phenomenal semantics
therefore first and foremost is an option interesting to the belief theorist.
As a response to the modularity phenomena, adopting a phenomenal semantics
brings doxastic accounts of experience back into business. Is this all it does? In a certain
sense, it is. The modularity phenomena do not establish one or the other of the
competing accounts as the true or correct one. Thus, there is a stand-off between the
two accounts.32 I think, however, that considering a phenomenal semantics for
experience with respect to the modularity phenomena nevertheless does something
towards undermining, or pre-empting, the motivations we could have for adopting a sui
generis account. Crucially, to have motivation for a sui generis account of the kind
suggested, we need to hold that there is no state of another kind that could play a
particular functional role. But sui generity with respect to experience becomes plausible
31
This was stressed to me by Tim Crane.
32
Here, too, I am indebted to Tim Crane.
29
only once functional role is characterized at the level of content – and on the assumption
that content is naive. Once phenomenal contents are so much as considered, it becomes
so much more difficult to have a non-question begging motivation for sui generity: To
have such a motivation, we would have to describe the relevant functional role
independently of the assumption of naive content. As long as we are concerned with the
modularity phenomena, these phenomena are precisely what characterize experience’s
functional role in a semantics-independent way. But if that is all that can non-question
beggingly be used to characterize experience’s functional role, then both naive sui
generis accounts and phenomenal doxastic accounts assign the very same functional
role to experience. It is thus not the case that no state of another kind could play the
particular role in question. Our motivation for adopting a sui generis account would
therefore seem to be seriously compromised.
Nevertheless, described on the level of content, the two accounts do induce different
functional roles for experience. The modularity phenomena just by themselves therefore
do not rule out a sui generis account of experience. Since both a phenomenal doxastic
and a naive sui generis account do equally well with respect to these phenomena, it is a
stand-off at this point. This changes, I think, once we consider another important
desideratum on a satisfactory account of experience: That of accommodating
experience’s reason-providing role. Here, I think a phenomenal belief account has a
clear advantage over any sui generis account. I shall consider this in the next, and final,
section.
4. Experience and Reasons
30
4.1 Folk-psychology and its Reasons
No-one, I take it, would deny the fundamental importance of perception to
epistemology. Perceptual experience plays a basic role in the justification of empirical
belief. Moreover, justificatory relations between experience and empirical belief are of
equally fundamental importance to the theory of content. Its justificatory relations to
experience arguably are among those relations that determine the very content of
empirical belief.
What is relevant here is only one particular aspect of the justificatory role of
experience for empirical belief, however.33 This is the aspect captured by the folkpsychological (and folk-epistemological) notion of reasons. Intuitively, this notion of a
reason is not restricted to theoretical reasoning; subjects can have reasons both for their
beliefs and their actions (or intentions). For simplicity’s sake, I shall focus on reasons
for belief in what follows. What we are interested in, then, are notions like those of
having a reason for belief or providing (a subject with) a reason for belief.
The notion of having a reason applies to subjects. It raises general questions as to
what kind of things reasons are, and what having them consists in, questions we don’t
33
Epistemologists offer us very different ways of thinking about justificatory relations. There are
externalist and internalist conceptions, and, these days, many people cook up more or less happy mixtures
of these. As far as this conflict concerns the final analysis of justification, or epistemic warrant, this is a
conflict we can stay out of. Whether or not epistemic warrant can, for instance, ultimately be analyzed in
terms of reliability does not matter to the following line of argument. This line of argument is thus not
hostage to the final outcome of any fundamental dispute in epistemology. Nor do we, by focusing on
reasons-relations, in any way have to reject externalist notions of warrant that are explicitly introduced as
alternatives to internalist notions of justification, and argued to obtain under specific circumstances (such
as, for instance, Burge’s notion of ‘perceptual entitlement’ (2003)).
31
have to take a stand on here. What we are concerned with is a particular case of having
reasons for belief: the case in which there is a proposition p that is the reason for
believing a proposition q, and the subject has this reason in the sense of having a
propositional attitude (of a certain kind) towards p. I shall apply the expression
‘provides (a subject with) a reason (for believing q)’ to the propositional attitude in
question, and reserve the term ‘reason’ for the propositional content of that attitude.34
Below, we shall hear more about both the kind of attitude capable of providing reasons
for belief, and the kind of relation between p and q required for such reason providing.
Intuitively, these are such that having a reason for a belief makes it in a certain sense
rational for the subject to have this belief.
The claim I shall base my line of argument on is this: It is an essential part of our
ordinary, everyday conception of experience that experience provides reasons for
empirical belief. Those of us capable of reflecting on and reporting our own reasons do,
on occasion, cite our experiences as reasons. This is a very simple, everyday business,
as McDowell nicely brings out:
[S]uppose one asks an ordinary subject why she holds some observational belief, say that
an object within her field of view is square. An unsurprising reply might be ‘Because it
looks that way’ (1994, 165).
Thus, I take it to be a desideratum on any satisfactory account of experience that it
accommodates their role as reason-providers in our folk-psychological practices of
explaining and justifying beliefs. We routinely cite experiences as reasons for what we
what we believe: I can, for instance, explain to you why I thought a certain tomato was
ripe by telling you that it looked red. This, we typically do when a belief has turned out
34
For this very practical terminology I am indebted to Marie Lundstedt.
32
to be false. But I can also use experience to argue for what I think is true. I can point to
the tomato and say: ‘It looks red. It’s ripe.’ Similarly, we use experiences to explain the
beliefs of others. This is both part of folk-psychology and of what we might call ‘folkepistemology’, our everyday practice of empirical justification.
The main assumption of what I shall call ‘the argument from reason’ can, thus, be
put like this:
(R)
Experiences provide subjects with reasons for first order empirical belief.35
The argument itself is very simple. It starts with (R) and contains only one more
premise:
(B)
The only propositional attitude that provides reasons for first order belief is that
of belief.36
35
Cf. Brewer, 1999, 18 for a very similar principle. In (1999) Brewer, however, gives a McDowell-
style ‘trancendental’ argument according to which something like (R) is a condition on having beliefs
with empirical content. What I am claiming is comparatively pedestrian; I am claiming that (R) is an
essential part of our pre-theoretic notion of experience, of folk-psychology and folk-epistemology and
should, therefore, be accommodated by any satisfactory account of experience. In (1999) and (2005)
Brewer also claims that only states with conceptual contents can provide reasons, and that subjects can
have reasons only if they can think of their reasons as reasons. I don’t see why any particular kind of
content would be required for a state to provide reasons (Byrne 2005 provides some instructive discussion
of this), and I do not think that reflection, or having the epistemic concepts of reason, justification etc. are
required for having reasons.
36
If you are reminded here of the (in)famous Davidsonian slogan: ‘Nothing can count as a reason for
holding a belief except another belief’ (1983, 310), that is fully intended. The claim I am making is,
however, not only formulated more carefully but also weaker. I am not making a claim about reasons in
33
The question, of course, is why we should subscribe to (B). To see this, we first need to
hear more about reasons. I shall therefore elaborate a bit on the notion of reasons on
which I think (B) holds. Because of its essential connections with folk-psychology, I
shall call this conception ‘the intuitive notion’; as we shall see, this notion is inferential
in nature.
Note, however, that I am not claiming that this is the only plausible or useful notion
of reason that there is. Rather, what I am claiming is this: The intuitive notion of
reasons is precisely that, a notion that is very intuitive. It is very intuitive because it
spells out what is basic to folk-psychology’s reasons-explanations or -justifications.
And folk-psychology routinely applies this notion to experiences. This does not exclude
the possibility that there are other useful notions of reason that also apply to
experiences.37 The claim is, however, that it is an essential part of our pretheoretic
notion of experience that it provides reasons in the sense spelled out here.
Consequently, the relevant desideratum cannot be satisfied by means of any other
notion of reason.
That said, here we go: What we are interested in are reasons for first order
(empirical) beliefs. Reasons explanations explain such beliefs by rationalizing them.
The rationality they provide is subjective in the sense of providing an explanation from
the subject’s own point of view, but the standards of rationality employed, though rather
general, but only about propositional attitudes providing subjects with reasons for first order beliefs.
Moreover, Davidson concludes that experiences (or at least sensations) are not reasons (cf. ibid.).
37
What I primarily have in mind here are (externalist, non-inferential) notions of entitlement. Both
Burge (2003) and McDowell (2004) argue that experience provides (externalist) entitlement to belief,
though of very different forms.
34
minimal in character, nevertheless are intersubjective, or objective, standards.38 Such
reasons-relations are content-sensitive: In order for one propositional attitude to provide
its subject with a reason for a (first order) belief, their contents need to be related in a
certain way. More precisely, where a propositional attitude towards p provides its
subject with a reason for believing that q, there is a relation of evidential or inferential
support between p and q. P is a reason for believing q, that is, only if p can be used as a
premise in a valid inference (logical or material) from p to q. Thus, the intuitive notion
of reason for (first order) belief is an inferential one. A propositional attitude towards p
provides its subject with a reason for believing q only if p evidentially supports q.39
Evidential support comes in degrees, and a subject S can have more than one reason
for believing q. Moreover, S can have both reasons for and against believing q. Having
a reason p for believing q does not imply that S actually believes that q. S might simply
never form the belief, or S’s reason for believing q might be overridden by reasons she
has against believing q. And even if S believes q, she might believe it for reasons other
38
These standards guarantee that the subject’s point of view brought out by such a rationalization is
one that others can recognize as a point of view, and that the rationalization provided has explanatory
power: It shows to others that, given the subject’s overall cognitive state, there was something that ‘spoke
for’ believing a certain proposition, something that would have ‘spoken for’ that belief for anyone being
in that overall cognitive state.
39
As I have characterized it, the intuitive conception of reasons is neutral on the question whether
reasons have to be true, or even known propositions, as required by Williamson (2000). Note that
Williamson nevertheless holds that perceptual illusions provide evidence, and that this is achieved by a
move similar to the one suggested here: ‘If perceptual evidence in the case of illusions consists of true
propositions, what are they? The obvious answer is: the proposition that things appear to be that way’
(2000, 198). Williamson does not explicitly construe this proposition as the content of the illusory
experience, however.
35
than p. The notion of reason we are concerned with here thus is the notion of a prima
facie reason.
For subjects to have such reasons, however, they do not need to be reflective
reasoners.40 In the basic, folk-psychological sense employed here, there are reasons as
soon as there are correct reasons explanations. Consequently, the reasons relations we
are interested in are relations between first order states. Such relations do not require
that their subjects draw any conscious inferences. Nor do they require that their subjects
need to be able to formulate their reasons, or think of them as reasons. Reflective
believers will normally be able to do so, but a subject does not need to be a reflective
believer to have reasons. Quite the contrary; having reasons (for first order beliefs) does
not require second order belief.
Let me summarize what I have said about the intuitive notion of reasons: This notion
concerns reasons for first order belief as provided by first order propositional attitudes.
Such reasons are prima facie, subjective reasons potentially rationalizing the relevant
beliefs (according to objective, if minimal, standards of rationality). A propositional
attitude towards p provides such reason for believing that q only if p evidentially or
40
Here, I agree with Burge who claims that requiring ‘reflection or understanding’ for having reasons
is ‘hyper-intellectualizing’ our cognitive lives (2003, 503ff, esp. fn 1), and that higher non-linguistic
animals do have reasons for some of their beliefs. Similarly, Dretske argues that certain non-linguistic
animals have reasons for their actions (cf. Dretske 2006). Burge and Dretske disagree on whether
experiences are among these reasons; Dretske holds that they are, while Burge denies this. According to
Burge, experiences only provide a different kind of warrant that he calls ‘entitlement’ (2003, 528).
McDowell (2004; 2008) also holds that experience does not provide reasons in the inferential sense, but
only what he calls ‘entitlement’ (which is rather different from Burgean entitlement). I am not concerned
here to dispute that experience provides entitlement in any of these senses; what I take exception to is
simply the counterintuitive contention that that is only kind of warrant they provide.
36
inferentially supports q. This is the notion of reason that folk-psychology routinely
applies not only to beliefs, but to experiences as well. When it comes to reasons for first
order empirical belief, folk-psychology is as happy with experiences as it is with
beliefs: it happily puts them into the same kind of ‘slot’ in theoretical reasoning. Thus,
both play exactly the same folk-psychological, reason-providing role.41
Now, given this notion of reason, we can next tackle the question: Why should we
subscribe to (B) above?
4.2 Reasons and Belief
Why does a propositional attitude have to be a belief in order to be a reason for belief in
the intuitive sense? It should be clear that an attitude providing reasons for belief cannot
be a conative attitude; desires, for instance, might provide practical reasons for actions
or intentions, but they do not provide reasons for belief. Next, none of what we might
call the ‘non-committal cognitive’ attitudes will do either. To see that, let’s assume that
q logically follows from p. And let’s start by considering mere tokenings of some
representation of p. No such ‘bare content’ could provide its subject with a reason to
believe anything. Of course, it is possible to merely entertain a proposition. But clearly,
such episodes do not figure as reasons in our psychology. That is the very point of
entertaining something – you just think it without actually drawing any consequences
from it. You might entertain the consequences, too, that is, you might entertain a
proposition p in order to see what the consequences would be were you to adopt it as
41
But is not belief also characterized by always also standing in need of reasons? This is a highly
controversial philosophical claim argued for by certain epistemologists of the coherentist stripe (most
influentially, the earlier BonJour (see, for instance, BonJour 1985, 78)), not part of the folk-psychology,
or even -epistemology, of belief.
37
one of your reasons. But merely entertaining p will not do the trick. And the same holds
for any other ‘non-commital’ attitude. Thus, you might imagine that p, or assume that p
(for instance, to prove that q follows from it), and none of these provide you with
reasons for actually endorsing q.
But what, then, makes an attitude into a reason-providing attitude? The answer, I
think, becomes clear if we consider what it is about these ‘non-committal’ attitudes that
disqualifies them: It is precisely that they do not commit to p. Assuming is particularly
instructive here: If you assume p in order to prove that q follows from p, what you
assume is the truth of p. But you just assume it, without actually committing to it. That
is precisely what prevents a mere assuming of p from providing you with a reason for
believing q – even if q logically follows from p, and even if you believe that it does. But
then, we can turn things around: In order to have such a reason, you precisely need to
commit to the truth of p – in other words: you need to hold p true.42 If you prefer to
think in terms of degrees of holding true, that is, of course, fine. But as far as I can see,
nothing else would do. There simply is no other kind of attitude that provides a subject
with reasons.43 On the assumption that any holding true is a belief, that gives us (B).
42
Of course, coming to believe that q follows from p might provide you with a reason for not
believing p, or even for believing not-p (for instance, if q is patently absurd). But our question was what
attitude towards p would make p into one or your reasons for further belief, not what you would have
reason to believe if you believed that q followed from p. So, if you believe that q follows from p, you
only have a reason for believing q if you (nevertheless continue to) hold p true.
43
This point is made specifically against McDowell in Stroud (2002, 89), XXX, and Ginsborg (2006).
Most recently, McDowell seems to have conceded the point: ‘If experiences have propositional content, it
is hard to deny that experiencing is taking things to be so, rather than what I want: a different kind of
thing that entitles us to take things to be so’ (2008, 15). McDowell now concludes that experiences do not
have propositional – but what he calls ‘intuitional’ – contents (I should note, however, that it is not
38
And the assumption that any holding true is a belief is extremely plausible; some might
say it is (almost) analytic.44 The notion of a propositional attitude in general, and that of
belief in particular, is a partly technical one essentially informed by our folkpsychological and folk-epistemological practices of ascribing, justifying, and predicting
both our own mental states and actions, and those of others. Because of this, the notion
of a propositional attitude is very austere: Propositional attitudes type psychological
states according to exactly two characteristics – attitude and content.45 Without austere
notions like these, notions subsuming a wide variety of otherwise quite different states,
a unifying model of folk-psychological reasoning would simply be impossible. The
notion of belief in particular is, in part, ‘designed’ to capture precisely the reasonproviding role of the relevant states. But then, to the extent that the notion of belief
completely clear to me whether McDowell’s intuitional contents do, or do not, fall under the wide notion
of a propositional content I have used in this paper).
44
Consequently, there is a tendency to think that it is not very informative to characterize belief in
terms of holding true; further analysis is needed to really understand the notion of belief. There is
currently no consensus on any suggested further analysis, however (for relevant discussion see, for
instance, Shah and Velleman 2005, Owens 2003, Wedgwood 2002, Zimmerman 2008). This is certainly
not the place to settle that issue; I hope the notion of holding true is intuitively clear enough for the
purposes of this paper. I take it that holding true does not include the attitude involved in imagining or
assuming. Actually holding a proposition p true among other things entails that a rational subject is
disposed to employ p as an asserted (as opposed to merely assumed) premise in both practical and
theoretical reasoning.
45
This does of course not mean that we cannot, for instance, type beliefs into all sorts of other sub-
kinds. Assume that experiences are beliefs, for instance. And assume that we can distinguish them from
other beliefs by their distinctive phenomenal character: There is something it is like to have an experience
that experiences do not share with any other beliefs. That makes experience into a particular species of
belief, but not into a different propositional attitude.
39
answers to the systematic needs of the theory of folk-psychology, a state essentially
playing the belief-role simply is a belief. And being a holding true is precisely what is
necessary and sufficient for being such a state. Therefore, (B) holds. The argument from
reasons concludes that experience is a kind of belief.
However, even if we agree that reason providing in the intuitive, inferential sense
requires holding true, we might at this point wonder whether the issue has not become
rather terminological.46 It might seem as if all participants to this debate easily could
agree that if we employ the austere, folk-psychological notion of belief their reasonproviding role makes experiences into a kind of belief. But this notion of belief might
not seem mandatory. Most importantly, the sui generis theorist does not seem to work
with this notion of belief; her notion of belief seems more narrow since it characterizes
belief by means of a functional role described at a level of more detail. Thus, our
disagreement with the sui generis theorist might seem to reduce to a mostly
terminological quarrel about what should properly be called a belief.
As I see it, this would be mistaken. What is at issue is, after all, folk-psychology, and
with it our everyday practices of ascribing, justifying, and predicting both our own
mental states and actions, and those of others. Capturing the concepts employed in these
practices more or less well can hardly be a merely terminological issue. But I shall not
press the issue any further in this way; rather, I shall try to bring out its substantial
nature by again comparing the doxastic account of experience I suggest with its main
competitor, the sui generis theory. For now, we are prepared to see how these accounts
do when it comes to accounting for the reason-providing role of experience.
46
This worry was brought home to me by an anonymous referee.
40
4.3 Stuttering Revisited
Prima facie, the argument from reasons provides strong support for a doxastic account
of experience. If experiences are beliefs, then we can account for their reason-providing
role on the model of beliefs providing reasons for other (first order) beliefs. So far,
however, we have ignored the last of the anti-doxastic arguments we looked at above:
the stuttering inference argument. This was precisely an argument to the conclusion that
any account according to which experience provides inferential reasons for empirical
belief results in stuttering inferences of the form p, therefore p. Which is absurd as a
model of the rational role of experience for empirical belief; experience clearly provides
strong, but defeasible inferential reasons, if any.
If the stuttering inference argument would actually establish this conclusion with the
intended generality, it would simply cancel whatever support for doxastic accounts is
provided by the argument from reasons. However, as we saw above, the stuttering
inference argument is premised on a naive semantics for experience. Stuttering is
avoided as soon as we work with an alternative semantics. The question then is whether
the alternative semantics in question induces a plausible picture of observational
reasoning. I shall argue now that in combination with a doxastic account a phenomenal
semantics does precisely that: On a phenomenal semantics, experience can be a kind of
belief and thus provide its subject with defeasible inferential reasons for further beliefs.
On the account suggested, experiences are beliefs about the way things look, sound, feel
or smell. Such beliefs provide precisely the right kind of support for basic perceptual
beliefs: That x looks red is a strong, but defeasible reason for believing that x is red. It is
a reason in precisely the same sense in which that there is smoke is a reason for
believing that there is fire. The relevant inference is material, not logical. This, it seems
to me, is precisely as it should be when it comes to experience providing defeasible
41
reasons for empirical belief.47 A phenomenal doxastic account of experience thus allows
us to account for the reason-providing role of experience by applying the intuitive,
inferential model of reasons for first order belief to experiences.
The next question is how the sui generis theory does with respect to reasonproviding: Can any such theory account for the reason-providing role of experience?
Before we get into that I would like to consider an objection that at least some sui
generis theorists might want to make here. The sui generis theorist might want to protest
that I have rigged the game against her from the start by selling the inferential element
as part and parcel of the intuitive, folk-psychological notion of a reason. After all, she
might say, there are other folk-psychologically respectable reasons relations which are
not inferential. Why should not the reasons relation between experience and empirical
belief be non-inferential, too?48
As I said above, I take it to be uncontroversial that first-order inter-doxastic reasons
relations are inferential. The dialectical upshot of the sui generis theorist’s protest would
therefore be the following: While it is indeed mandatory for the belief theorist to use the
inferential model of reasons, this might not be so for the sui generis theorist. As long as
she can provide us with a model on which experiences provide reasons in an intuitive,
but non-inferential sense there is no reason to think her account inferior. I think this is
47
Thanks to an anonymous referee for prompting clarification here.
48
Pryor, for instance, inveighs against what the calls the ‘Premise Principle’, insisting, among other
things, that it is not the case that all reasons are propositions (cf. Pryor 2005, 189ff). However, none of
Pryor’s arguments concern the case we are solely concerned with here: The case where a propositional
attitude provides its subject with reasons for first-order belief. In this case, it seems to me, it is very hard
indeed to understand how an attitude with content p could provide a reason for believing q where p does
not inferentially support q.
42
wrong; there is reason to think that any account that cannot apply the inferential model
of reasons to experience is inferior to one that can. For even if there were relevant noninferential reasons-relations, folk-psychology certainly does not mark any such
difference between experience and belief. Intuitively, experience and beliefs provide
reasons in the very same sense.
Moreover, there is no model around we could plausibly use here, nothing, at least,
that would come even close to being as well-worked out and well-understood as the
inferential model. And the question is whether there even could be any model on which
experiences provide reasons in an intuitive, but non-inferential sense. What are the
candidates for folk-psychological, but non-inferential reasons relations? There are the
practical reasons desires or other pro-attitudes provide for actions or intentions. Whether
these relations are inferential or not is a matter of debate, but it should be immediately
clear that the (epistemic) reasons experiences provide for beliefs cannot instructively be
modeled on the motivations desires provide for actions. Next, one might think that first
order beliefs (and other mental states) provide reasons for certain second order beliefs
their subject might form about them. Whether these really are reasons is a matter of
debate, but if they are, the relation is clearly non-inferential. Again, however, it should
be immediately clear that the reasons experience provides for first-order belief cannot
be modeled on those first order states provide for second order belief; there is no sense
in which first order empirical beliefs are about our experiences. As far as I can see, that
exhausts the possibilities. There is thus no reason to think that experience’s reason-
43
providing role can be modelled on any other non-inferential folk-psychological reasons
relation.49
What is left for the sui generis theorist is to simply declare, as a matter of
dogmatism, that experience does provide (non-inferential) reasons for empirical first
order belief.50 The sui generis theorist thus seems to be driven to the claim that
experiences not only are sui generis states, but also provide sui generis reasons. But
why should we believe that? In the absence of any other kind of reasons relation to use
as a model, some substantial further explanations are required before we can so much as
assume that anything recognizable as a reasons relation has been identified.51 As long
as these are not forthcoming, the inferential model arguably remains the only, and
49
An idea here might be to go outside folk-psychology; experience’s reason-providing role could for
instance be modelled on testimony (Evans (1982, 123) hints at this, and Burge’s (2003, 452) contention
that experiences are commitments, but not of the ‘whole animal’ could be construed along this lines, too).
Since we are concerned with personal level propositional attitudes, attitudes of the experiencing subject
herself, some rather implausible level of compartmentalization would seem to arise in known illusion
cases, but the idea might still seem worthwhile to explore. Be that as it may, the most that can be drawn
from this at the moment is an open cheque on the future.
50
Cf. Pryor 2000. Pryor writes: ‘My view is that whenever you have an experience as of p, you
thereby have immediate prima facie justification for believing p’ (536), and he explicitly characterizes
this view of perceptual justification as ‘dogmatism’ (532).
51
Pryor only hints at such an explanation in a footnote. After dismissing, rightly to my mind, possible
explanations in terms of irresistibility and conceptual truth, he suggests that is is ‘the peculiar
“phenomenal force” or way our experiences have of presenting propositions to us’ (547, fn. 37) that
explains why experiences provide justification. If this is a reference to the so-called ‘transparency’ of
experience, as Pace surmises, or to those aspects of ‘perceptual immediacy’ that have to do with vividness
etc., it remains unclear what epistemic significance such phenomenal features of experience could have
(cf. Pace 2008, 21ff).
44
certainly the best understood and most well-established conception of reasons for first
order empirical belief available. Even for those not convinced that folk-psychology’s
egalitarian treatment of experience and belief forces this model on experiences anyway,
this should be a very good reason for favoring accounts of experience able to use the
inferential model.52
The next question, however, is whether it is an advantage that cannot equally be
claimed by the sui generis theorist. After all, the sui generis theorist agrees with us that
experiences are strongly representational states, i.e. that they represent the world as
being such that their contents are true. Thus, it might seem as if the sui generis theorist
could, so to speak, buy the first half of the argument from reason: He could accept that
all strongly representational states involve the attitude of holding true, but, it might
seem, he could also insist that that is not enough for making a strongly representational
state into a belief (in his more narrow sense). Consequently, the inferential model of
reasons would seem to be as available to the sui generis theorist as to the belief theorist.
And thus, we would have reached a theoretical stand-off again.
This time, however, the impression of stand-off is mistaken. The stuttering inference
argument applies as soon as we conceive of experiences as providing inferential
52
Someone who clearly appreciates this advantage is Heck, who writes:
I take it to be a well-established, and familiar, point that perceptions are not beliefs.
This is unfortunate, for if they were, we would have a relatively easy answer to the
question how experience justifies beliefs about the world. (...) Though one can
certainly raise questions about how some beliefs justify other beliefs (how the
beliefs we now hold give us reasons to hold other ones), these sorts of questions
seem relatively tractable – much more tractable, anyway, than questions about how
perceptions justify beliefs. It is just in the nature of beliefs to stand in justificatory
relations with other beliefs (...) (2000, 507f).
45
reasons. Whether the reason providing attitude is one of belief or sui generis does not
matter here. As long as the sui generis theorist works with a naive semantics, he is stuck
with stuttering, and thus with a very implausible picture of observational reasoning. A
phenomenal belief theory thus clearly has a decisive advantage over a naive sui generis
account.
This is not the end of the matter, however. As we saw above, there is no reason in
principle why the sui generis theorist could not adopt a phenomenal semantics, too. As
long as we were concerned with accounting for the modularity phenomena only, there
was no real motivation for such a move. But now, stuttering seems to provide such a
motivation. And then, the question would be whether a phenomenal doxastic account
has any advantage over a phenomenal sui generis theory. The answer to this question is
that once both work with a phenomenal semantics, the doxastic account has every
advantage over the sui generis account: There is no level of description left at which the
functional role assigned to experiences by these two accounts is different.
At the end of section 3, we compared naive sui generis accounts with phenomenal
doxastic accounts and concluded that the motivation for the former was weakened
because there was no non-question begging level of description left at which functional
roles differed. At the level of content, the functional roles assigned to experiences by
the competing accounts nevertheless remained different.
The situation is different now: Once we construe experiences as states with
phenomenal contents, the peculiarity of functional role induced by naive contents
simply vanishes. Whether experiences with phenomenal contents are construed as
beliefs or as sui generis attitudes ceases to make any functional difference whatsoever.
Thus, it is not only the case that we do not have much reason to think that experiences
46
are sui generis, they simply cannot be: We have lost the difference that was to make the
distinction.
The conclusion of these considerations therefore must be that even though a naive sui
generis account and a phenomenal doxastic account of experience do equally well when
it comes to accounting for the modularity phenomena, the doxastic account has the
advantage when it comes to another desideratum: In contrast to any sui generis account,
the phenomenal doxastic account neatly accommodates the reason providing role of
experience. As far as the desiderata here considered are concerned, the belief theory
thus is not only back as a viable contender to an account of experience, it also has a
decisive advantage over its strongest competitor.53
Department of Philosophy
Stockholm University
53
Obviously, the considerations presented amount to no more than a partial defense of a phenomenal
belief-theory of experience. Issues that need to be adressed elsewhere include animals and small children,
and hallucinations. Two brief remarks: Concerning animals and small children, I think that as long as we
do not require them to possess the relevant concepts (in any sense in which that would be impossible for
them), there is no obstacle to ascribing phenomenal contents to their experiences. Rather, this would seem
to be required, and justified, by precisely the thought that their experiences have the same contents, and
are (at least capable of) playing the same folk-psychological role as ours do. Regarding hallucinations, on
pain of a known hallucination argument, a belief theory has to construe the content of hallucinations as
free from existential commitment, be it incurred by singular or existentially quantified general contents.
Otherwise the knowing hallucinator holds contradictory beliefs. My preferred solution, pace
disjunctivism, aims at avoiding such contradiction while holding on to the idea that veridical experiences
and the corresponding hallucinations have the same content. The basic idea is to construe ‘looks’
primarily as a sentential operator L1 (L1(p); ‘It looks as if p’), and to define a predicate modifier L2
(L2(F)(x); ‘x looks F’) in terms of it: ∃x ((L2(F)(x)) ≡df ∃x L1(Fx).
47
References
Armstrong, D. W. 1968: A Materialist Theory of the Mind. London: Routledge.
BonJour, L. 1985: The Structure of Empirical Knowledge. Cambridge, MA: Harvard
University Press.
Brewer, B. 1999: Perception and Reason. Oxford: Clarendon Press.
Brewer, B. 2005: Do sense experiential states have conceptual content? In E. Sosa and
M. Steup (eds), Contemporary Debates in Epistemology. Oxford: Blackwell.
Brewer, B. 2008: Perception and content. http://www2.warwick.ac.uk/fac/soc/philosophy/staff/brewer/pc.ejp.pdf.
Broad, C. D. 1923: Scientific Thought. London : Kegan Paul, Trench, Trubner.
Burge, T. 2003: Perceptual entitlement. Philosophy and Phenomenological Research
LXVII, 503-48.
Byrne, A. 2005: Perception and conceptual content. In M. Steup and E. Sosa (eds.),
Contemporary Debates in Epistemology. Oxford: Blackwell.
Campbell, J. 2002: Berkeley’s puzzle. In T. Szabo Gendler and J. Hawthorne (eds.),
Conceivability and Possibility. Oxford: Clarendon Press.
Chisholm, R. 1957: Perceiving: A Philosophical Study. Ithaca, NY: Cornell University
Press.
Craig, E. 1976: Sensory experience and the foundations of knowledge. Synthese 33: 124.
Crane, T. 1988: The waterfall illusion. Repr. in Y. Gunter (ed.), Essays on
Nonconceptual Content. Cambridge, MA: MIT (Bradford) 2003.
48
Crane, T. 1992: The nonconceptual content of experience. In T. Crane (ed.), The
Contents of Experience. Cambridge: Cambridge University Press.
Crane, T. 2001: Elements of Mind. Oxford: Oxford University Press.
Crane, T. 2005: The problem of perception. The Stanford Encyclopedia of Philosophy
(Spring
2008
Edition),
Edward
N.
Zalta
(ed.),
http://plato.stanford.edu/-
archives/spr2008/entries/perception-problem/.
Davidson, D. 1983: A coherence theory of truth and knowledge. In E. Lepore (ed.),
Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson.
Oxford: Blackwell 1986.
Davies, M. 1992: Perceptual content and local supervenience. Proceedings of the
Aristotelian Society 92, 21-45.
Dretske, F. 1981: Knowledge and the Flow of Information. Cambridge, MA: MIT Press.
Dretske, F. 1995: Naturalizing the mind. Cambridge, MA: MIT Press.
Dretske, F. 2006: Perception without awareness. In T. Szabo Gendler and J. Hawthorne
(eds), Perceptual Experience. Oxford: Clarendon Press.
Evans, G. 1982: The Varieties of Reference. Oxford: Clarendon Press.
Ginsborg, H. 2006: Reasons for belief. Philosophy and Phenomenological Research 72,
286-318.
Glüer, K. 2004: On perceiving that. Theoria 70, 197-212.
Hamlyn, D. W. 1994: Perception, sensation and nonconceptual content. Repr. In Y.
Gunther (ed.), Essays on Nonconceptual Content. Cambridge, MA: MIT (Bradford)
2003.
Harman, G. 1990: The intrinsic quality of experience. Philosophical Perspectives 4, 3152.
49
Heck, R. G. 2000: Non-Conceptual content and the ‘space of reasons’. Philosophical
Review 109, 483-523.
Jackson, F. 1977: Perception. Cambridge: Cambridge University Press.
Martin, M. G. F. 2002: The transparency of experience. Mind and Language 17, 376425.
McDowell, J. 1994: Mind and World. Cambridge, MA: Harvard University Press.
McDowell, J. 1997: Brandom on representation and inference. Philosophy and
Phenomenological Research 57, 157-62.
McDowell, J. 1998: Reply to commentators. Philosophy and Phenomenological
Research 58, 403-31.
McDowell, J. 2002: Replies. In N. Smith (ed.), Reading McDowell on ‘Mind and
World’. London: Routledge.
McDowell, J. 2004: Reply to Kathrin Glüer. Theoria LXX, 213-15.
Owens, D. J. 2003: Does belief have an aim? Philosophical Studies 115, 283-305.
Pace, M. 2008: Perceptual knowledge and the metaphysics of experience. Forthcoming
in The Philosophical Quarterly, published online: 22-Feb-2008, doi: 10.1111/j.14679213.2008.536.x.
Peacocke, C. 1992: A Study of Concepts. Cambridge, MA: MIT Press.
Pitcher, G. 1971: A Theory of Perception. New Jersey: Princeton University Press.
Pryor, J. 2005: There is immediate justification. In M. Steup and E. Sosa (eds),
Contemporary Debates in Epistemology. Oxford: Blackwell.
Searle, J. 1983: Intentionality. Cambridge: Cambridge University Press.
Sellars, W. 1963: Science, Perception and Reality. London: Routledge.
Shah, N. and Velleman, D. 2005: Doxastic deliberation. Philosophical Review 114, 497534.
50
Shoemaker, S. 1994: Phenomenal character. Nous 28, 21-39.
Shoemaker, S. 2000: Introspection and phenomenal character. Philosophical Topics 28,
247-73.
Shoemaker, S. 2007: On the way things appear. In T. Szabó Gendler and J. Hawthorne
(eds), Perceptual Experience. Oxford: Oxford University Press.
Siegel, S. 2005: The contents of perception. The Stanford Encyclopedia of Philosophy
(Spring
2008
Edition),
Edward
N.
Zalta
(ed.),
http://plato.stanford.edu/-
archives/spr2008/entries/perception-contents/.
Speaks, J. 2005: Is there a problem about nonconceptual content? Philosophical Review
114, 359-398.
Stroud, B. 2002: Sense-experience and the grounding of thought. In N. Smith (ed.),
Reading McDowell. London: Routledge.
Travis, C. 2004: The silence of the senses. Mind 113, 57-94.
Tye, M. 2002: Consciousness, Color, and Content. Cambridge, MA: MIT Press.
Wedgwood, R. 2002: The Aim of Belief. Philosophical Perspectives 16, 276-297.
Williamson, T. 2000: Knowledge and its Limits. Oxford: Oxford University Press.
Wright, C. 1998: McDowell’s oscillation. Philosophy and Phenomenological Research
58, 395-402.
Zimmerman, A. 2008: The nature of belief. Forthcoming in Consciousness Studies.
11990 words.
51